PROJECT SUMMARY/ABSTRACT Experimental animal and clinical testing to evaluate hepatotoxicity demands extensive resources and long turnaround times. Utilization of computational models to directly predict the toxicity of new compounds is a promising strategy to reduce the cost of drug development and to screen the multitude of industrial chemicals and environmental contaminants currently lacking safety assessments. However, the current computational models for complex toxicity endpoints, such as hepatotoxicity, are not reliable for screening new compounds and face numerous challenges. Our recent studies have shown that traditional Quantitative Structure-Activity Relationship modeling is applicable for relatively simple properties or toxicity endpoints with a clear mechanism, but fails to address complex bioactivities such as hepatotoxicity. The primary objective of this proposal is to develop novel mechanism-driven Virtual Adverse Outcome Pathway (vAOP) models for the fast and accurate assessment of hepatotoxicity in a high-throughput manner The resulting vAOP models will be experimentally validated using a complement of in vitro and ex vivo testing. We have generated a preliminary vAOP model based on the antioxidant response element (ARE) pathway that has undergone initial validation and refinement using in vitro testing. To this end, our project will generate novel predictive models for hepatotoxicity by applying 1) a virtual cellular stress pathway model to mechanism profiling and assessment of new compounds; 2) computational predictions to fill in the missing data for specific targets within the pathway; 3) in vitro experimental validation with three complementary bioassays; and 4) ex vivo experimental validation with pooled primary human hepatocytes capable of biochemical transformation. The scientific approach of this study is to develop a universal modeling workflow that can take advantage of all available short-term testing information, obtained from both computational predictions using novel machine learning approaches and in vitro experiments, for target compounds of interest. We will validate and use our modeling workflow to directly evaluate the hepatotoxicity of new compounds and prioritize candidates for validation in pooled primary human hepatocytes. The resulting workflow will be disseminated via a web portal for public users around the world with internet access. Importantly, this study will pave the way for the next generation of chemical toxicity assessment by reconstructing the modeling process through a combination of big data, computational modeling, and low cost in vitro experiments. To the best of our knowledge, the implementation of this project will lead to the first publicly available mechanisms-driven modeling and web- based prediction framework for complex chemical toxicity based on publicly-accessible big data. These deliverables will have a significant public health impact by not only prioritizing compounds for safety testing or new chemical development, but also revealing toxicity mechanisms.